Configuration
The Alquimia Platform runtime is configured through environment variables grouped into logical areas: runtime mode, authentication, storage, secrets, telemetry, and feature flags. This page explains the model; exact variable names and defaults live in the runtime repository.
Configuration areas
Section titled “Configuration areas”Runtime mode
Section titled “Runtime mode”The runtime can run as a master, a worker, or both in a single process:
| Mode | Use |
|---|---|
| Master | HTTP API and registry management |
| Worker | Consumes events and executes agents |
| All | Both roles in one process; useful for local development |
Authentication
Section titled “Authentication”The runtime supports several authentication methods:
- API tokens for service-to-service calls.
- JWT validated against a Keycloak or compatible OIDC issuer.
- OIDC through Keycloak for user-facing endpoints.
See Authentication & authorization for the conceptual model.
Storage backends
Section titled “Storage backends”| Backend | Purpose | Common options |
|---|---|---|
| PostgreSQL | Persistent audit worklog and registry | Managed cloud Postgres |
| Redis | Session state, distributed locks, caches | Redis, KeyDB |
| Qdrant | Vector search for knowledge bases | Qdrant Cloud, self-hosted |
| S3 / MinIO | Blob storage for attachments | AWS S3, MinIO, Ceph |
Secrets
Section titled “Secrets”Secrets are resolved at runtime from HashiCorp Vault or environment variables. Vault supports dynamic credentials for PostgreSQL and Redis, as well as static secrets for API keys, S3 credentials, and webhook signing keys.
Telemetry
Section titled “Telemetry”OpenTelemetry is configured through standard OTLP endpoints for metrics, traces, and logs. See Observability for how the signals correlate.
Best practices
Section titled “Best practices”- Keep secrets out of
.envfiles in production; use Vault or your cluster’s secret store. - Run master and worker in separate deployments for independent scaling.
- Set resource requests and limits on both roles.
- Enable telemetry endpoints before going live so you have baseline metrics.